This PR fixes an inconsistency in get_table_dataframe where empty datasets would bypass their specific reconstruction logic if storage.blocks is empty( for eg when you call recorder.clear()). Because Datasets do not store time in their metadata, this premature short-circuit caused their empty DataFrame outputs to be missing the time column entirely.

Changes:

• Removed the global short-circuit for empty storage.blocks.
• Empty datasets now correctly dispatch to their specific converters
• The global short-circuit is now strictly reserved for completely fresh/uninitialized datasets (data_type == unknown).

Previously if you run

model = BoltzmannWealth()
model.run_for(5)
model.recorder.clear()

print(model.recorder.get_table_dataframe("agent_data"))
print(model.recorder.get_table_dataframe("model_data"))

Old output

Empty DataFrame
Columns: [unique_id, wealth]
Index: []
Empty DataFrame
Columns: [gini, time]
Index: []

Note: Here time is in columns because modeldataset stores data as a dictionary( for eg. {"gini":.., "time":..) but this behavior would be changed once #3389 is merged

New Output

Empty DataFrame
Columns: [unique_id, wealth, time]
Index: []
Empty DataFrame
Columns: [gini, time]
Index: []

If you run above setup with start_time > 5( i.e data_type=="unknown") then output would be

Old output

Empty DataFrame
Columns: []
Index: []
Empty DataFrame
Columns: []
Index: []

New Output

Empty DataFrame
Columns: [time]
Index: []
Empty DataFrame
Columns: [time]
Index: []